<![CDATA[Stress is a serious challenge for students that can negatively impact physical health, mental well-being, and academic performance. However, accurate and effective stress detection approaches to support early intervention are still limited. This study aims to evaluate machine learning models for detecting student stress levels with optimal accuracy to facilitate early intervention. The research employs a quantitative approach using a dataset containing 1,100 student samples from Nepal, encompassing 20 stress-related features from psychological, social, academic, environmental, and physiological aspects. Data were collected via a self-report questionnaire, processed with StandardScaler scaling, and analyzed using 10-fold cross-validation. The models tested include Perceptron, Gradient Boosting Trees Classifier (GBTC), Naive Bayes (NB), Logistic Regression (LR), and AdaBoost. The results show that Perceptron performed the best with an accuracy of 97.27%, followed by NB (95.45%), GBTC (94.54%), LR (94.54%), and AdaBoost (93.63%). Perceptron, with its advantage in linearity and evaluation through 10-fold cross-validation, shows great potential as an effective classification model for student stress detection, which can accelerate early intervention and enhance student well-being and learning environments.]]>
Copyrights © 2024
